A liquid leakage range prediction method and system

Abstract

This invention discloses a method and system for predicting the extent of liquid leaks. The method first responds to abnormal events involving hazardous chemical transport vehicles by simultaneously acquiring visible light video, thermal imaging video, and lidar point cloud data. Then, based on the event location, an initial analysis area is defined, and the area boundary is dynamically corrected based on a 3D bridge surface model constructed by lidar and the flow tendency identified by visible light video. Next, a temperature change rate map is generated within the dynamic area through temporal difference calculation, and superpixel regions are extracted using a clustering algorithm. When the area and temperature change rate exceed an adaptive threshold, a preliminary leak detection signal is generated. Finally, the leak source is located by tracing back the temporal data, and the liquid gravity diffusion path is simulated using the bridge surface slope field to predict the future contamination range. This invention achieves early and accurate identification of leaks and scientific prediction of contamination range, significantly improving the efficiency and reliability of emergency response to hazardous chemical leaks on bridge surfaces.

Description

Technical Field

[0001] This invention belongs to the field of liquid positioning technology, and in particular relates to a method and system for predicting the range of liquid leakage. Background Technology

[0002] With the continuous improvement of the highway network, accidents involving the transport of hazardous chemicals and their leakage occur frequently. Especially on bridge sections crossing sensitive water bodies such as drinking water source protection areas and Class II or above surface water bodies, improper handling of leaked hazardous chemicals can lead to pollution spread and cause serious ecological and environmental impacts.

[0003] Currently, the handling of hazardous chemical leaks on provincial highway bridges mainly relies on bridge deck runoff pipes for collection and emergency storage in accident pools. However, the identification, collection, and disposal of hazardous chemical leaks primarily depend on manual judgment, resulting in a low level of automation and intelligence. Existing technologies have the following shortcomings:

[0004] Recognition lag: Traditional manual recognition methods are slow to react and cannot achieve early warning;

[0005] Inaccurate location: There is a lack of accurate means to locate the leak source, making it difficult to quickly determine the extent of contamination;

[0006] Insufficient predictive capability: Unable to effectively predict liquid diffusion paths and future contamination extent;

[0007] Insufficient utilization of multi-source data: failure to effectively integrate multimodal data such as visible light, thermal imaging, and lidar;

[0008] Although existing technologies have developed monitoring methods based on single sensors, they all suffer from problems such as limited monitoring range, high false alarm rate, and poor positioning accuracy. For example, relying solely on thermal imaging technology may result in false alarms due to changes in ambient temperature; and using only visible light video is difficult to work effectively at night or in low visibility conditions. Summary of the Invention

[0009] The present invention aims to overcome the shortcomings of the prior art and provide a method and system that can quickly and accurately predict the extent of liquid leakage, thereby achieving early warning, precise location and intelligent emergency response.

[0010] In a first aspect, the present invention provides a method for predicting the extent of liquid leakage, comprising:

[0011] In response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge, the system acquires visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event.

[0012] Based on the spatial location of the abnormal event, a first analysis region is initially defined in the thermal imaging video stream;

[0013] Based on the lidar point cloud data, a 3D model of the bridge deck terrain is created to obtain a 3D model of the bridge deck. Based on the 3D model of the bridge deck and the liquid flow tendency identified by the visible light video stream, the boundary of the first analysis area is dynamically corrected to generate a second dynamic analysis area that matches the potential leakage path.

[0014] Within the second dynamic analysis region, temporal difference calculation is performed on the thermal imaging video stream of consecutive frames to generate a temperature change rate map. A clustering algorithm is then used to extract candidate superpixel regions with similar change characteristics from the temperature change rate map. When the area and temperature change rate of the candidate superpixel region both exceed the adaptive threshold, a leakage initial judgment signal is generated.

[0015] In response to the initial leak detection signal, based on the time-series data of the temperature change rate spectrum, the center of the connected domain that first showed a significant temperature anomaly and was identified as having the same liquid characteristics is determined as the leak source point. Combined with the three-dimensional model of the bridge deck, the slope field is calculated to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

[0016] Secondly, the present invention provides a liquid leakage range prediction system, comprising:

[0017] The acquisition module is configured to acquire visible light video stream, thermal imaging video stream and lidar point cloud data under the same spatiotemporal reference as the abnormal event in response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge surface.

[0018] The segmentation module is configured to initially delineate a first analysis region in the thermal imaging video stream based on the spatial location of the abnormal event.

[0019] The correction module is configured to perform a three-dimensional modeling of the bridge deck terrain based on the lidar point cloud data to obtain a three-dimensional model of the bridge deck, and dynamically correct the boundary of the first analysis area based on the three-dimensional model of the bridge deck and the liquid overflow tendency identified by the visible light video stream to generate a second dynamic analysis area that matches the potential leakage path.

[0020] The generation module is configured to perform temporal difference calculation on the thermal imaging video stream of consecutive frames within the second dynamic analysis region to generate a temperature change rate map, and extract candidate superpixel regions with similar change characteristics from the temperature change rate map using a clustering algorithm. When the area and temperature change rate of the candidate superpixel regions both exceed an adaptive threshold, a leakage initial judgment signal is generated.

[0021] The prediction module is configured to respond to the initial leak detection signal, and based on the time-series data of the temperature change rate spectrum, determine the center of the connected domain that first shows a significant temperature anomaly and is identified as having the same liquid characteristics as the leak source point, and combine the bridge deck three-dimensional model to calculate the slope field, simulate the diffusion path of the liquid under gravity, and predict the pollution range in the future.

[0022] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the liquid leakage range prediction method according to any embodiment of the present invention.

[0023] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the liquid leakage range prediction method of any embodiment of the present invention.

[0024] The liquid leakage range prediction method and system of this application employs a first analysis region anchored by the locations of previously identified abnormal events. Through a dual mechanism of "fixed radius ensuring basic coverage + vehicle size-driven dynamic adjustment," it accurately defines the minimum sufficient observation range centered on the hazardous chemical vehicle itself within the thermal imaging data stream. This not only ensures that critical leakage points such as tank valves and pipeline interfaces are included in the monitoring field without omission, but also, by discarding a large amount of irrelevant bridge surface background areas, the second dynamic analysis region integrates a centimeter-precision 3D bridge surface model constructed by lidar with real-time diffuse characteristics captured by visible light video, shifting the focus of analysis from the "vehicle itself." The system intelligently evolves to "liquid behavior," generating a dynamic monitoring corridor that closely matches the actual diffusion path of leaked liquid constrained by gravity and terrain through slope-driven physical flow direction simulation and visual flow trace verification. This regional evolution strategy, from "point" to "line" and then to "surface," enables the system to proactively predict the movement direction of the leaked material. This not only minimizes the risk of missed detection caused by the liquid flowing out of the first analysis area but also keeps computational resources focused on the highest probability pollution diffusion channels. As a result, the system can lock onto the development trend in the very early stages after a leak occurs, laying the spatial perception foundation for subsequent accurate positioning and range prediction. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating a method for predicting the extent of liquid leakage according to an embodiment of the present invention;

[0027] Figure 2 This is a structural block diagram of a liquid leakage range prediction system provided in an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figure 1 The diagram shows a flowchart of a liquid leakage range prediction method according to this application.

[0031] like Figure 1 As shown, the method for predicting the extent of a liquid leak includes the following steps:

[0032] Step S101: In response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge, acquire visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event.

[0033] In this step, predefined abnormal events include, but are not limited to: vehicle rollover: determined by analyzing whether the angle between the vehicle's bounding box and the road surface horizontal line lasts for more than 30 degrees for more than 500 milliseconds. This abnormal event identification task is performed in real-time by edge computing nodes (such as NVIDIA Jetson AGX Orin) deployed on the bridge side. Optimized YOLOv8 models run on these nodes for vehicle detection and preliminary behavior analysis.

[0034] Time synchronization:

[0035] Master clock source: The system is equipped with a high-precision GPS / BeiDou timing module, which serves as the PTP master clock for the entire sensor network;

[0036] Synchronization Protocol: All sensors (cameras, thermal imagers, LiDAR) and edge computing nodes support the IEEE 1588 PTP protocol. During system initialization, the accuracy of each device and the master clock is calibrated to ensure that the time synchronization error of the entire system is less than 1 millisecond.

[0037] Data stamping: Each frame of video data (visible light, thermal imaging) and each lidar point cloud data packet is marked with a precise timestamp provided by the PTP protocol during acquisition.

[0038] Spatial synchronization (registration) scheme:

[0039] Intrinsic and extrinsic parameter calibration: During the system deployment phase, calibration objects such as checkerboard grids are used to jointly calibrate the visible light camera and the thermal imaging camera to accurately obtain their intrinsic parameters (focal length, principal point, distortion coefficient) and extrinsic parameters (rotation matrix and translation vector), thereby achieving pixel-level spatial alignment.

[0040] Radar-camera joint calibration: By using a specific calibration board, the transformation relationship between the lidar coordinate system and the visible light camera coordinate system is established, realizing the fusion of point cloud data and image data;

[0041] Online spatial reference: All sensor data is ultimately converted to a world coordinate system with a fixed corner point in the 3D model of the bridge deck as the origin. The lidar point cloud directly provides 3D coordinates, and the video images, through the aforementioned calibration parameters and coordinate transformations, can map the pixels in the images to the world coordinate system.

[0042] The edge computing node sends a global trigger signal and broadcasts it via gigabit industrial Ethernet.

[0043] Upon receiving the global trigger signal, each sensor immediately performs the following operations:

[0044] Cache backtracking: Due to data processing delays, the system instructs all sensors to retrieve data from their respective circular buffers, starting 100 milliseconds before the trigger and continuing to collect data afterward. This ensures that complete data is captured at the moment the event occurs.

[0045] Data encapsulation: Bind data frames (video frames / point cloud packets) with spatiotemporal stamps to the event ID in the trigger signal to form standardized data units;

[0046] Transmission: The packaged data units are transmitted in real time to the cloud analysis platform or local server via a high-speed network for in-depth analysis in subsequent steps S102 to S105.

[0047] Step S102: Based on the spatial location of the abnormal event, initially delineate a first analysis region in the thermal imaging video stream.

[0048] In this step, the spatial location of the abnormal event is taken as the center point; an initial circular area is defined with a preset fixed radius; the radius of the initial circular area is dynamically adjusted according to the outer rectangle size of the hazardous chemical transport vehicle obtained from the visible light video stream based on the preset target recognition model, so that the adjusted area can completely cover the outer rectangle of the vehicle, and the adjusted area is determined as the first analysis area.

[0049] In this embodiment, a global trigger signal is received from step S101, which contains the precise spatial location of the abnormal event (i.e., the X, Y, Z coordinates in the world coordinate system).

[0050] The three-dimensional spatial position is projected onto the image pixel coordinate system of the thermal imaging video stream using a pre-calibrated coordinate transformation matrix to obtain the corresponding pixel coordinates (u, v). These coordinates are the center point C of the initial circular region.

[0051] A fixed radius R1 is preset, based on empirical data, to ensure coverage of the initial impact range of common small leaks. For example, R1 can be set to 2 meters. A circular area is delineated on the thermal imaging image with center point C as the center and R1 as the radius.

[0052] At the same time, a preset target recognition model (such as an optimized YOLOv8s model) is invoked to process the video frames in the visible light video stream at the trigger time.

[0053] The target recognition model identifies the hazardous materials transport vehicle that triggered the incident and outputs its circumscribed rectangle. The length L and width W (in pixels) of this circumscribed rectangle are obtained, and the radius R2 of the circumscribed circle is calculated. The radius R2 can be calculated as half the length of the rectangle's diagonal. To ensure the analysis area completely covers the vehicle, the dynamic radius is set to: ,in, This is a safety margin used to ensure that the area boundary is slightly larger than the vehicle outline.

[0054] Using the original center point C as the center and the calculated dynamic radius R as the new radius, a new circular region is delineated on the thermal imaging image. This new circular region is the final first analysis region.

[0055] In summary, this step initializes the system with a fixed radius centered on the event location, ensuring rapid response and basic coverage, and providing a stable initial analysis anchor point. Secondly, the actual vehicle dimensions are dynamically adjusted to adjust the radius, allowing the analysis area to adapt to different specifications of hazardous chemical transport vehicles. This effectively avoids the problem of missing key leakage points on the vehicle body (such as tank valves and pipeline interfaces) due to an excessively small area, or introducing too much irrelevant background interference due to an excessively large area. Finally, this strategy, while ensuring the completeness of leakage identification, greatly reduces the computational load of subsequent image processing algorithms by precisely focusing computational resources on the most relevant areas.

[0056] Step S103: Based on the lidar point cloud data, perform three-dimensional modeling of the bridge deck terrain to obtain a three-dimensional model of the bridge deck. Based on the three-dimensional model of the bridge deck and the liquid overflow tendency identified by the visible light video stream, dynamically correct the boundary of the first analysis area to generate a second dynamic analysis area that matches the potential leakage path.

[0057] In this step, the lidar point cloud data is preprocessed, including denoising and ground point cloud segmentation; the segmented ground point cloud is reconstructed into a continuous triangular mesh surface using the Poisson surface reconstruction algorithm or the Delaunay triangulation algorithm; the triangular mesh surface is used as the three-dimensional model of the bridge deck, which contains the three-dimensional coordinate information of each point on the bridge deck.

[0058] In one specific implementation, a plane fitting algorithm is used to fit a reference plane representing the bridge deck from the denoised point cloud. A distance threshold (e.g., 5 cm) is set, and all points within this threshold are segmented and identified as ground point clouds. The ground point cloud is converted into an octree structure, and the implicit surface is reconstructed by solving the Poisson equation. Finally, isosurfaces are extracted to generate a watertight triangular mesh surface. Alternatively, the ground point cloud is subjected to 3D Delaunay triangulation, and then, through filtering and optimization, the set of triangles constituting the bridge deck terrain is retained. The reconstructed triangular mesh surface is the 3D model of the bridge deck. Each vertex contains its 3D coordinates (X, Y, Z), and each patch consists of three vertex indices.

[0059] Furthermore, slope field and aspect data are extracted from the 3D model of the bridge deck; based on the spatial location of abnormal events, combined with slope field and aspect data, the initial flow direction of the liquid under gravity is simulated; the first analysis region is extended downstream along the simulated initial flow direction; inter-frame difference or optical flow calculations are performed on the visible light video stream to identify potential liquid overflow regions where brightness and texture change; the extended region and the potential liquid overflow region are combined and convex hull processing is performed to form the second dynamic analysis region.

[0060] In a specific implementation process, the normal vector of each triangular facet in the 3D model of the bridge deck is calculated. The slope is defined as the angle between the normal vector and the zenith direction (positive Z-axis direction); the aspect is defined as the projection direction of the normal vector onto the horizontal plane (XY plane). Through mesh calculation, the slope field and aspect field covering the entire bridge deck are obtained.

[0061] The spatial location of the abnormal event is projected onto the three-dimensional model of the bridge deck, serving as the starting point P for the simulation.

[0062] Simulated flow direction: Based on the aspect field, starting from the starting point P, the initial path of the water flow is simulated along the steepest downhill direction (i.e., the aspect). This can be achieved using the D8 single-flow-direction algorithm, which calculates the maximum slope direction between the central grid and its 8 neighboring grids as the flow direction.

[0063] Extension and Expansion: The first analysis region is extended downstream along the simulated initial flow path. The extension distance L is positively correlated with the average bridge deck slope S, for example, L = base extension length × (1 + S × coefficient). This forms an initial extension region.

[0064] For visible light video streams, two or three consecutive frames are taken, converted to grayscale, and then differential operations are performed. The differential results are thresholded to obtain the moving target region. Noise is removed by morphological opening operations, and holes are filled by closing operations. Finally, the visual flow region with significant brightness changes is extracted.

[0065] The initially expanded region and the visually diffused region are logically ORed to merge them into a single region. The convex hull of the merged region's outer boundary point set is then calculated using either the Graham Scan algorithm or the QuickHull algorithm. This step aims to eliminate unnecessary depressions and complex contours within the region, forming a complete and coherent second dynamic analysis region.

[0066] Step S104: Within the second dynamic analysis region, perform temporal difference calculation on the thermal imaging video stream of consecutive frames to generate a temperature change rate map, and extract candidate superpixel regions with similar change characteristics from the temperature change rate map using a clustering algorithm. When the area and temperature change rate of the candidate superpixel regions both exceed the adaptive threshold, generate a leakage preliminary judgment signal.

[0067] In this step, three consecutive thermal imaging images are selected; for each pixel in the second dynamic analysis area, the temperature change rate of each pixel between the first two frames and the last two frames is calculated; the calculated temperature change rate values ​​of all pixels are combined into a matrix corresponding to the pixel coordinates of the second dynamic analysis area, i.e., the temperature change rate map.

[0068] In one specific implementation, three consecutive frames containing the second dynamic analysis region are extracted from the thermal imaging video stream: the thermal imaging image at time t-1. Thermal imaging image at time t And thermal imaging images at time t+1 ;

[0069] For each pixel (i, j) within the second dynamic analysis region, extract its temperature value over three consecutive frames: the temperature value of pixel (i, j) in the thermal imaging image at time t-1. The temperature value of pixel (i, j) in the thermal image at time t. The temperature value of pixel (i, j) in the thermal image at time t+1. ;

[0070] The temperature change rate of pixel (i, j) is calculated using the central difference method: , For time intervals;

[0071] Temperature change rate of all pixels This is combined into a two-dimensional matrix that corresponds one-to-one with the pixel coordinates of the second dynamic analysis region. This two-dimensional matrix is ​​the temperature change rate map. This graph visually illustrates the degree of drastic temperature changes at various points within a specific time window.

[0072] A linear iterative clustering algorithm is used to perform superpixel segmentation on the temperature change rate map, dividing the image into multiple superpixel regions with similar temperature change rates. The average and variance of the temperature change rate of all pixels in each superpixel region are calculated. Superpixel regions with an average value exceeding a primary threshold and a variance below a specific threshold are preliminarily identified as candidate superpixel regions with similar change characteristics.

[0073] In another specific implementation, the Simple Linear Iterative Clustering (SLIC) algorithm was used to analyze the temperature change rate map. The system is segmented to obtain superpixel regions. Specifically, the SLIC algorithm segments the temperature change rate map in (x, y, ...) Clustering is performed in three-dimensional space to generate K superpixels. The pixels within each superpixel have high similarity in spatial location and temperature change rate. The value of K is set according to the area of ​​the second dynamic analysis region.

[0074] For each segmented superpixel region, the average and variance of the temperature change rate of all pixels within it are calculated. Superpixels that satisfy the condition that the average value > the average value threshold and the variance < the variance threshold are initially identified as candidate superpixel regions. This means that a significant and consistent temperature change has occurred in this region.

[0075] The adaptive threshold is dynamically adjusted based on the ambient background temperature. The adaptive threshold includes a minimum area threshold and a temperature threshold. It is determined whether the physical area of ​​each candidate superpixel region is greater than the minimum area threshold and whether the average temperature change rate of each candidate superpixel region is greater than the current temperature threshold under the current ambient background temperature. If the physical area of ​​each candidate superpixel region is greater than the minimum area threshold and the average temperature change rate of each candidate superpixel region is greater than the current temperature threshold under the current ambient background temperature, then a leakage initial judgment signal is generated.

[0076] In another specific implementation process, the minimum area threshold Amin is set according to actual needs and is used to filter out irrelevant areas that are too small, for example, Amin = 0.02 square meters.

[0077] Temperature threshold This is a dynamic threshold, adjusted based on the ambient background temperature. Adaptive adjustment is performed. The adjustment formula is as follows:

[0078] ,

[0079] In the formula, The reference temperature threshold is, for example, 5.0°C. For compensation coefficient, The reference ambient temperature is, for example, 20°C.

[0080] Step S105: In response to the initial leakage judgment signal, based on the time series data of the temperature change rate spectrum, the center of the connected domain that first shows a significant temperature anomaly and is identified as having the same liquid characteristics is determined as the leakage source point. Combined with the three-dimensional model of the bridge deck, the slope field is calculated to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

[0081] In this step, starting from the moment the initial leakage judgment signal is generated, the time series data of the temperature change rate spectrum is traced backward; the set of pixels where the earliest temperature change rate exceeds the adaptive threshold is found; connectivity analysis is performed on the set of pixels to identify the earliest connected component; the earliest geometric center of the earliest connected component is calculated, and the projection coordinates of the earliest geometric center on the 3D model of the bridge deck are determined as the leakage source point.

[0082] In this embodiment, starting from the moment the initial leakage judgment signal is generated, the historical temperature change rate spectrum sequence stored at that moment is traced back to the beginning.

[0083] Traverse each frame of the historical temperature change rate map in chronological order from morning to night. For each frame of the historical temperature change rate map, perform binarization processing using the same dynamic temperature threshold as in step S104 to obtain a binary image Bt, where a pixel with a value of 1 represents a temperature change rate exceeding the threshold.

[0084] When the number of pixels with a value of 1 in Bt first exceeds the minimum cluster threshold (e.g., 10 consecutive pixels), backtracking stops, and the time of this frame is the estimated time of the leak start.

[0085] A connected component labeling algorithm (such as the Two-Pass algorithm) is applied to the binary image B0 corresponding to the estimated leakage start time to identify all connected pixel sets. Among these connected components, the one with the largest physical area is selected as the earliest connected component (based on the assumption of area maximization: the initial leakage point will form the most significant continuous region).

[0086] Calculate the earliest geometric center of the earliest connected component (i.e., the average of all pixel coordinates), denoted as pixel P0. Using camera calibration parameters and coordinate transformation, back-project P0 onto the 3D model of the bridge deck to obtain its 3D coordinates in the world coordinate system. This coordinate is the final determined source of the leak.

[0087] The slope and aspect of each grid cell are calculated from the 3D model of the bridge deck. Starting from the leakage source, a pre-set particle system model is used to simulate the flow process of liquid under the combined action of gravity and road friction based on the slope and aspect of each grid cell. The prediction time is set, the particle system model is run, and all grid cells that the liquid may cover within the prediction time are obtained. The set of all grid cells is output as the predicted pollution range for the future time.

[0088] In this embodiment, the three-dimensional model of the bridge deck (triangular mesh) is uniformly rasterized into grid cells with a fixed resolution (e.g., 0.1m × 0.1m). For each grid cell, the slope value and aspect (expressed in angles) of the cell are calculated based on the normal vector of the triangular facet to which it belongs.

[0089] Using a pre-defined particle system model, the flow process of liquid under the combined effects of gravity and road surface friction is simulated as follows:

[0090] To leak the source Centered on the prediction area, N particles (e.g., N=1000) are randomly initialized, each particle representing a tiny liquid unit.

[0091] Sports rules:

[0092] Dominant force: Gravity, particle size at each simulation time step Within 0.1 seconds, it moves along the slope of its grid cell;

[0093] Movement speed: Where S is the slope, g is the gravitational acceleration, R is the hydraulic radius (which can be simplified by using the thickness of the liquid film represented by the particle), and C is the coefficient of friction (set according to the road surface material). For square root operations;

[0094] Random perturbation: A small random angular perturbation (e.g., ±10°) is introduced to simulate the uncertainty of liquid diffusion and the microscopic unevenness of the road surface.

[0095] Boundary conditions:

[0096] Bridge edge boundary: When a particle moves to the edge of the bridge, it is marked as "outflowed" and removed from the simulation;

[0097] Guardrail boundary: When a particle collides with the guardrail, its direction of motion changes according to the preset reflection rules.

[0098] Set a prediction time T (e.g., 5 minutes), run the particle system model until the simulation time reaches T, and count the grid cells covered by all particles that did not flow out of the bridge surface and were not removed within the time T. The set of these grid cells is the predicted contamination range.

[0099] In summary, the method of this application uses the locations of previously identified abnormal events as anchor points in the first analysis region. Through a dual mechanism of "fixed radius ensuring basic coverage + vehicle size-driven dynamic adjustment," it accurately defines the minimum sufficient observation range centered on the hazardous chemical vehicle itself in the thermal imaging data stream. This not only ensures that key leakage points such as tank valves and pipeline interfaces are included in the monitoring field of view without any omissions, but also intelligently shifts the focus of analysis from the "vehicle itself" by discarding a large amount of irrelevant bridge surface background areas. The second dynamic analysis region integrates the centimeter-precision 3D bridge surface model constructed by lidar with the real-time flow characteristics captured by visible light video, thus eliminating the need for background detection. Moving to the "liquid behavior" stage, a dynamic monitoring corridor is generated through slope-driven physical flow simulation and visual flow trace verification, which highly matches the actual diffusion path of the leaked liquid constrained by gravity and terrain. This regional evolution strategy from "point" to "line" and then to "area" enables the system to proactively predict the movement direction of the leaked material. This not only minimizes the risk of missed detection caused by the liquid flowing out of the first analysis area, but also keeps computational resources focused on the highest probability pollution diffusion channel. As a result, the development trend can be locked in at a very early stage after the leak occurs, laying the spatial perception foundation for subsequent accurate positioning and range prediction.

[0100] Please see Figure 2 The diagram shows a structural block diagram of a liquid leak range prediction system according to this application.

[0101] like Figure 2 As shown, the liquid leakage range prediction system 200 includes an acquisition module 210, a division module 220, a correction module 230, a generation module 240, and a prediction module 250.

[0102] The acquisition module 210 is configured to acquire visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event in response to a predefined abnormal event occurring on the bridge surface transport vehicle; the segmentation module 220 is configured to initially delineate a first analysis region in the thermal imaging video stream based on the spatial location of the abnormal event; the correction module 230 is configured to perform three-dimensional modeling of the bridge surface terrain based on the lidar point cloud data to obtain a three-dimensional model of the bridge surface, and dynamically correct the boundary of the first analysis region based on the three-dimensional model of the bridge surface and the liquid overflow tendency identified by the visible light video stream, generating a second dynamic analysis region that matches the potential leakage path; the generation module 24... 0, configured to perform temporal difference calculation on the thermal imaging video stream of consecutive frames within the second dynamic analysis area to generate a temperature change rate map, and extract candidate superpixel regions with similar change characteristics from the temperature change rate map using a clustering algorithm. When the area and temperature change rate of the candidate superpixel regions both exceed an adaptive threshold, a preliminary leakage judgment signal is generated. The prediction module 250, configured to respond to the preliminary leakage judgment signal, determine the center of the connected region that first shows a significant temperature anomaly and is identified as having the same liquid characteristics as the leakage source point based on the temporal data of the temperature change rate map, and calculate the slope field in conjunction with the three-dimensional model of the bridge deck to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

[0103] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0104] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the liquid leakage range prediction method in any of the above method embodiments.

[0105] In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows:

[0106] In response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge, the system acquires visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event.

[0107] Based on the spatial location of the abnormal event, a first analysis region is initially defined in the thermal imaging video stream;

[0108] Based on the lidar point cloud data, a 3D model of the bridge deck terrain is created to obtain a 3D model of the bridge deck. Based on the 3D model of the bridge deck and the liquid flow tendency identified by the visible light video stream, the boundary of the first analysis area is dynamically corrected to generate a second dynamic analysis area that matches the potential leakage path.

[0109] Within the second dynamic analysis region, temporal difference calculation is performed on the thermal imaging video stream of consecutive frames to generate a temperature change rate map. A clustering algorithm is then used to extract candidate superpixel regions with similar change characteristics from the temperature change rate map. When the area and temperature change rate of the candidate superpixel region both exceed the adaptive threshold, a leakage initial judgment signal is generated.

[0110] In response to the initial leak detection signal, based on the time-series data of the temperature change rate spectrum, the center of the connected domain that first showed a significant temperature anomaly and was identified as having the same liquid characteristics is determined as the leak source point. Combined with the three-dimensional model of the bridge deck, the slope field is calculated to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

[0111] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the liquid leak extent prediction system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely disposed relative to a processor, which can be connected to the liquid leak extent prediction system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0112] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the liquid leak range prediction method of the above-described method embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the liquid leak range prediction system. The output device 340 may include a display screen or other display device.

[0113] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0114] In one implementation, the above-described electronic device is applied to a liquid leak range prediction system for a client, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0115] In response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge, the system acquires visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event.

[0116] Based on the spatial location of the abnormal event, a first analysis region is initially defined in the thermal imaging video stream;

[0117] Based on the lidar point cloud data, a 3D model of the bridge deck terrain is created to obtain a 3D model of the bridge deck. Based on the 3D model of the bridge deck and the liquid flow tendency identified by the visible light video stream, the boundary of the first analysis area is dynamically corrected to generate a second dynamic analysis area that matches the potential leakage path.

[0118] Within the second dynamic analysis region, temporal difference calculation is performed on the thermal imaging video stream of consecutive frames to generate a temperature change rate map. A clustering algorithm is then used to extract candidate superpixel regions with similar change characteristics from the temperature change rate map. When the area and temperature change rate of the candidate superpixel region both exceed the adaptive threshold, a leakage initial judgment signal is generated.

[0119] In response to the initial leak detection signal, based on the time-series data of the temperature change rate spectrum, the center of the connected domain that first showed a significant temperature anomaly and was identified as having the same liquid characteristics is determined as the leak source point. Combined with the three-dimensional model of the bridge deck, the slope field is calculated to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

[0120] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the extent of liquid leakage, characterized in that, include: In response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge, the system acquires visible light video stream, thermal imaging video stream, and lidar point cloud data under the same spatiotemporal reference as the abnormal event. Based on the spatial location of the abnormal event, a first analysis region is initially defined in the thermal imaging video stream; Based on the lidar point cloud data, a 3D model of the bridge deck terrain is created to obtain a 3D model of the bridge deck. Based on this 3D model and the liquid flow tendency identified by the visible light video stream, the boundary of the first analysis area is dynamically corrected to generate a second dynamic analysis area that matches the potential leakage path, including: Slope field and aspect data were extracted from the three-dimensional model of the bridge deck. Based on the spatial location of the abnormal event, and combined with the slope field and aspect data, the initial flow direction of the liquid under gravity is simulated. The first analysis region is extended downstream along the simulated initial flow direction; Perform inter-frame difference or optical flow calculations on the visible light video stream to identify potential liquid flow regions where brightness and texture change; The extended region is combined with the potential liquid overflow region, and convex hull processing is performed to form the second dynamic analysis region; Within the second dynamic analysis region, temporal difference calculation is performed on the thermal imaging video stream of consecutive frames to generate a temperature change rate map. A clustering algorithm is then used to extract candidate superpixel regions with similar change characteristics from the temperature change rate map. When the area and temperature change rate of the candidate superpixel region both exceed the adaptive threshold, a leakage initial judgment signal is generated. In response to the initial leak detection signal, based on the time-series data of the temperature change rate spectrum, the center of the connected domain that first showed a significant temperature anomaly and was identified as having the same liquid characteristics is determined as the leak source point. Combined with the three-dimensional model of the bridge deck, the slope field is calculated to simulate the diffusion path of the liquid under gravity and predict the pollution range in the future.

2. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, The initial delineation of the first analysis region in the thermal imaging video stream based on the spatial location of the abnormal event includes: The spatial location of the abnormal event is taken as the center point; Define an initial circular region with a preset fixed radius; Based on the outer rectangle size of the hazardous chemical transport vehicle obtained from the visible light video stream using a preset target recognition model, the radius of the initial circular region is dynamically adjusted so that the adjusted region can completely cover the outer rectangle of the vehicle, and the adjusted region is determined as the first analysis region.

3. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, The step of performing three-dimensional modeling of the bridge deck terrain based on the lidar point cloud data to obtain the three-dimensional bridge deck model includes: The lidar point cloud data is preprocessed, including noise reduction and ground point cloud segmentation; Using the Poisson surface reconstruction algorithm or the Delaunay triangulation algorithm, the segmented ground point cloud is reconstructed into a continuous triangular mesh surface; The triangular mesh surface is used as the three-dimensional model of the bridge deck, and the three-dimensional model of the bridge deck contains the three-dimensional coordinate information of each point on the bridge deck.

4. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, Within the second dynamic analysis region, performing temporal difference calculations on the continuous frames of the thermal imaging video stream to generate a temperature change rate map includes: Select three consecutive thermal imaging images; For each pixel in the second dynamic analysis region, calculate the rate of temperature change of each pixel between the first two frames and the last two frames. The calculated temperature change rate values ​​of all pixels are combined into a matrix corresponding to the pixel coordinates of the second dynamic analysis area, which is the temperature change rate map.

5. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, The step of extracting candidate superpixel regions with similar change characteristics from the temperature change rate map using a clustering algorithm includes: The temperature change rate map is segmented into superpixel regions using a linear iterative clustering algorithm, dividing the image into multiple superpixel regions with similar temperature change rates. Calculate the mean and variance of the rate of temperature change for all pixels within each superpixel region; Superpixel regions whose average value exceeds a primary threshold and whose variance is below a specific threshold are initially identified as candidate superpixel regions with similar variation characteristics.

6. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, When both the area and temperature change rate of the candidate superpixel region exceed the adaptive threshold, the generation of the initial leakage judgment signal includes: The adaptive threshold is dynamically adjusted based on the ambient background temperature, and the adaptive threshold includes a minimum area threshold and a temperature threshold. Determine whether the physical area of ​​each candidate superpixel region is greater than the minimum area threshold, and determine whether the average temperature change rate of each candidate superpixel region is greater than the current temperature threshold under the current ambient background temperature; If the physical area of ​​each candidate superpixel region is greater than the minimum area threshold, and the average temperature change rate of each candidate superpixel region is greater than the current temperature threshold under the current ambient background temperature, then the initial leakage judgment signal is generated.

7. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, The step of identifying the center of the connected region where the earliest significant temperature anomaly was observed and which was identified as having the same liquid characteristics, based on the time-series data of the temperature change rate spectrum, as the leakage source point includes: Starting from the moment the initial leakage detection signal is generated, the time-series data of the temperature change rate spectrum is traced backward; Find the set of pixels where the earliest temperature change rate exceeds the adaptive threshold; Perform connectivity analysis on the set of pixels to identify the earliest connected components; Calculate the earliest geometric center of the earliest connected component, and determine the leakage source point by projecting the earliest geometric center onto the three-dimensional model of the bridge deck.

8. The method for predicting the extent of liquid leakage according to claim 1, characterized in that, The calculation of the slope field using the three-dimensional model of the bridge deck, simulating the diffusion path of the liquid under gravity, and predicting the pollution range in the future include: The slope and aspect of each grid cell are calculated from the three-dimensional model of the bridge deck. Starting from the leakage source, based on the slope and aspect of each grid cell, a preset particle system model is used to simulate the flow process of liquid under the combined action of gravity and road friction. Set a prediction time, run the particle system model, and obtain all grid cells covered by liquid within the prediction time; The set of all grid cells is output as the predicted pollution range for the future time period.

9. A liquid leakage range prediction system, characterized in that, include: The acquisition module is configured to acquire visible light video stream, thermal imaging video stream and lidar point cloud data under the same spatiotemporal reference as the abnormal event in response to a predefined abnormal event occurring on a hazardous chemical transport vehicle on the bridge surface. The segmentation module is configured to initially delineate a first analysis region in the thermal imaging video stream based on the spatial location of the abnormal event. The correction module is configured to perform 3D modeling of the bridge deck terrain based on the lidar point cloud data to obtain a 3D model of the bridge deck. Based on the 3D model of the bridge deck and the liquid flow tendency identified by the visible light video stream, it dynamically corrects the boundary of the first analysis area and generates a second dynamic analysis area that matches the potential leakage path, including: Slope field and aspect data were extracted from the three-dimensional model of the bridge deck. Based on the spatial location of the abnormal event, and combined with the slope field and aspect data, the initial flow direction of the liquid under gravity is simulated. The first analysis region is extended downstream along the simulated initial flow direction; Perform inter-frame difference or optical flow calculations on the visible light video stream to identify potential liquid flow regions where brightness and texture change; The extended region is combined with the potential liquid overflow region, and convex hull processing is performed to form the second dynamic analysis region; The generation module is configured to perform temporal difference calculation on the thermal imaging video stream of consecutive frames within the second dynamic analysis region to generate a temperature change rate map, and extract candidate superpixel regions with similar change characteristics from the temperature change rate map using a clustering algorithm. When the area and temperature change rate of the candidate superpixel regions both exceed an adaptive threshold, a leakage initial judgment signal is generated. The prediction module is configured to respond to the initial leak detection signal, and based on the time-series data of the temperature change rate spectrum, determine the center of the connected domain that first shows a significant temperature anomaly and is identified as having the same liquid characteristics as the leak source point, and combine the bridge deck three-dimensional model to calculate the slope field, simulate the diffusion path of the liquid under gravity, and predict the pollution range in the future.

Images